Springer

http://www.jstor.org/stable/20116795 .

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Springer is collaborating with JSTOR to digitize, preserve and extend access to Synthese.

http://www.jstor.org

This content downloaded from 195.220.213.10 on Tue, 16 Apr 2013 05:40:15 AM

All use subject to JSTOR Terms and Conditions
 ROBERTO MAIOCCHI

PIERRE DUHEM'S THE AIM AND STRUCTURE

OF PHYSICAL THEORY:
A BOOK AGAINST CONVENTIONALISM

ABSTRACT. I reject the widely held view that Duhem's 1906 book La Th?orie physique
is a statement of instrumentalistic conventionalism, motivated by the scientific crisis at
the end of the nineteenth century. By considering Duhem's historical context I show that
his epistemological were
views already formed before the crisis occured; that he consis
tently supported general thermodynamics against the new atomism; and that he rejected
the epistemological views of the latter's philosophical supporters. In particular I show
that Duhem rejected Poincar?'s account of scientific language, Le Roy's view that laws
are definitions, and the conventionalist's use of simplicity as the criterion of theory choice.
Duhem regarded most theory choices as decidable on empirical grounds, but made
historical context the main determining factor in scientific change.

Duhem's famous book La Th?orie physique is almost universally con

sidered one of the most significant documents of that cultural movement
addressed against positivist optimism. Reflecting on the crisis of nine
teenth-century mechanism, at the beginning of our century, this move
ment generated an instrumentalistic conception of scientific knowledge.
Duhem's text has always been considered one of the most brilliant and
vital - perhaps the most vital - of the conventionalist movement, the
skeptical, philosophical answer to the difficulties of classical science.
The study of Duhem's intellectual biography (Maiocchi 1985) has led
me to reach conclusions in many ways diametrically opposed to tra
ditional judgments. These may be synthesized in a formula which is
only apparently paradoxical: the main intent of the Th?orie physique
was to oppose instrumentalism, subjectivism, and the devaluation of
the cognitive power of science.
The first observation to be made, apparently a point of chronology,
but of decisive importance, is that the epistemological theses contained
in the Th?orie physique of 1906 were clearly and fully expressed by
Duhem in a series of articles written between 1892 and 1894 (Duhem
1892a, 1892b, 1893a, 1893b, 1893c, and 1894a). Thus Duhem's
1892c,
epistemology predates the discovery of radioactive phenomena, Gouy's
experiments on Brownian motion, and Kaufmann's experiments on the
variable mass of the electron, as well as the introduction of quantum

Synthese 83: 385-400, 1990.

? 1990 Kluwer Academic Publishers. Printed in the Netherlands.

This content downloaded from 195.220.213.10 on Tue, 16 Apr 2013 05:40:15 AM

All use subject to JSTOR Terms and Conditions
386 ROBERTO MAIOCCHI

hypothesis and the first relativistic hypotheses. In short, Duhem pre

sented his epistemological theses before the "undoing of all principles",
to use Poincar?'s ill-timed expression, and before the explosion of the
'crisis' of the sciences.
In fact, the thematics of crisis are totally absent in Duhem. On the
contrary, all of Duhem's historical and epistemological reflections, all
of his scientific work, as a researcher reveals the conviction that science
was not only undergoing a period of great splendor during the late
nineteenth century, but was getting rid of the errors that had ac
companied it through the last three centuries! Duhem's criticism of
mechanism never attacks the trust in mechanics as a theory of mathe
matical physics, but always and only rejects attempts to extend mechan
ics into a nonscientific, metaphysical sphere. Above all, his criticism of
mechanism is based upon the fundamental assumption that there is a
better theory than rational mechanics, i.e., generalized thermodynam
ics. It is the success of thermodynamics that imposes the necessity of
constructing a new mechanics, not the failure of the old one. Thanks
to the new generalized mechanics, the dreams of the boldest mechan
ists, such as Berthollet, seemed to be on the verge of coming true.
The new mechanics did not reject classical mechanics, but enlarged
and generalized it. Classical mechanics became the model for rigor as
well as for method (Newton's, obviously, not Descarte's), and for the
form given to one's own principles, which were required to maintain
the closest possible analogy with the classical ones. The new mechanics
stayed close to its classical model, rather than opposing it, following
the original program of energetics formulated by Rankine and carried
on by the mechanist William Thomson, before the latter became what
Duhem called a "modelist" (see Duhem 1893d).
In Duhem's opinion, the developments in nineteenth century science
confirm the positivist belief in a continuous progress of scientific knowl
edge from other methodological bases: "In our days, many are being
swept by a wave of skepticism", but those who force themselves to find
in science "the continuation of a tradition of a slow but steady pro
gress", will see "that a theory that disappears, never disappears com
pletely" (Duhem 1894b, p. 124). It is not the crisis of science, but its
successes which impose upon Duhem the necessity of epistemological
reflection.
Duhem's interpretation of scientific theories as simple instruments of
classification does not appear as an answer to a supposed crisis of

This content downloaded from 195.220.213.10 on Tue, 16 Apr 2013 05:40:15 AM

All use subject to JSTOR Terms and Conditions
 A BOOK AGAINST CONVENTIONALISM 387

mechanism. Not is his epistemology

only free of such a 'crisis', but such
instrumentalistic conceptions had already been present in the French
milieu for several decades. Diffidence toward hypotheses, a phenom
enalist view of science, and an instrumentalistic, manipulative interpre
tation of theories were widely diffused ideas, and dominant among the
French scientists of the positivist age. The French scientists' ideal was
personified by the likes of Regnault, Bertin, Berthelot, Sainte-Claire
- all
De ville, Jamin, Cornu, Violle, and Le Chatelier fundamentally
experimenters. They sporadically showed an ideological belief in the
supreme value of mechanics, although in fact they produced 'anti
modelist' physics. In many cases, their work was mathematically poor,
deaf to the calls of theoretical physics and insensitive to nuances of
experiment. Even more clearly than contemporary physics, the milieu
of chemistry, in which Duhem was trained, showed a general mistrust
of the idea that scientific theories might yield explanations, in the
sense of revealing hidden truths behind phenomena. These objections
appeared most clearly in the case of the atomic theory. Atomism was

interpreted in the first place as a classifying tool, even by its supporters

like Wurtz. Duhem's understanding of science
and epistemology was
fundamentally influenced by Henri
Sainte-Claire Deville. In the work
of this French chemist-physicist, Duhem found, even before Mach, the
very clearly stated idea that every scientific theory is simply a classifying
tool (see Sainte-Claire Deville 1866).
Unlike Mach in Germany, Duhem did not have to fight against
dogmatic belief in the nascent cognitive power of mechanics or against
a tendency to objectify models. The cognitive devaluation of theories
and models was already extensively employed as a criticism by French
positivism. As criticism it was not particularly discerning, but certainly
historically effective. Duhem had to fight a battle in exactly the opposite
direction: contrary to the flattening out imposed by the empiricist
method of his predecessors, he had to avenge the rights of theory,
showing how the ineliminable theoretical components present in every
observation gave
meaning to the scientist's experimental work.
Duhem's epistemology was a defense of theories against positivist pre
tenses to eliminate them by strictly reducing science to pure experience.
The positivists considered theories as secondary tools when compared
to experience, even as superfluous and therefore, eliminable. Duhem
endeavored to show that theories are the heart of a scientific venture.

This content downloaded from 195.220.213.10 on Tue, 16 Apr 2013 05:40:15 AM

All use subject to JSTOR Terms and Conditions
388 ROBERTO MAIOCCHI

His was a radically anti-inductivist, antiempiricistic epistemology, a

'praise of theories', and in that sense it was opposed to Mach's.
Even Duhem's most renowned battle against Anglo-Saxon physics
cannot be understood as a fight against the identification of models and
reality, denying cognitive value to the models. Duhem never identified,
but rather, he always underscored the distinction between Cartesian
mechanism and that of Kelvin and Maxwell: in British mechanism,
Duhem immediately recognized the model had only heuristic functions;
itwas a working tool for the physicist who needed to 'illustrate', satisfy
ing the need of concrete interpretation (see Duhem 1893b). The funda
mental charge against the British was not the fact that they used models,
but their use of models in an incoherent way, conceiving them precisely
as simple instruments. Even Laplacian physics was 'modelist', and yet,
on repeated occasions, Duhem referred to it as one of the great ex
amples of a theory in physics. Anglo-Saxon 'modelism', propagated in
France by Poincar?, seemed to Duhem the most dangerous variant of
instrumentalism, and he fought it by taking an explicitly anti-instrumen
talistic position: if theories were simple classifying tools, it would be
perfectly normal to adopt various criteria for different classifications,
introducing incoherence in physics by using different models to repre
sent the same object, as the British did. Incoherence, (i.e., British
physics as supported by Poincar?) can be fought only by admitting that
theories are classifying tools, being neither arbitrary nor subjective, but
leaning rather not toward the construction of a 'natural classification',
namely, one having objective significance.
Theoretical coherence obsessed Duhem's research. He sought the
rigorous structuring of scientific terms in a deductive, hypothetical sys
tem which conceded nothing, in matters of rigor, to intuition or com
mon sense. Generalized thermodynamics was the perfect answer to
these requirements (and where it fell short, Duhem took great pains
to make the necessary corrections). Anglo-Saxon 'modelism' instead
proposed an uncoordinated physics, a gallery of images that, due to
lack of coherence, could not be judged a theory. But coherence was
sustainable and justifiable only by admitting that theories, inasmuch as
they are constructed to organize mathematically the world of phenom
ena, are also capable of reflecting an ever-perfectible and always 'more
perfect' real arrangement, rather than a subjective one. We know with
certainty
-
according to Duhem in 1893 - that relations among material

This content downloaded from 195.220.213.10 on Tue, 16 Apr 2013 05:40:15 AM

All use subject to JSTOR Terms and Conditions
 A BOOK AGAINST CONVENTIONALISM 389

substances are "neither undetermined nor contradictory". Therefore,

when faced with physics proposing two irreconcilable theories:

We are certain that the classification is not in conformity

proposed by such physics with
the natural order of 'the laws . . .making the incoherence disappear, we will have some

probability of bringing it closer to that order to make it more natural, thus, more
perfect_(Duhem 1893c, pp. 369-70)

The idea of 'natural classification' was judged an extrinsic ideological

addition, and even contradictory with Duhem's epistemology. Yet this
idea was enunciated from 1893 as the methodological axis carrying the
fight against British physics and in favor of generalized thermodynam
ics. Without it, all of Duhem's scientific work would be meaningless.
Not only that, but the whole of his epistemology and his historical work
was an effort to sustain this notion. The pivotal problem around which
all of the Th?orie physique hinges is just that: how to reconcile an
unprejudiced, pitiless and extremely acute critique of the scientist's
work with the idea of a science that has cognitive value. How does one
criticize the dogmatic empiricism of positivism without falling into the
subjectivity of instrumentalism? In order to understand why this prob
lem had become so important in Duhem's eyes during the first few
years of the new century, we should remember the genesis of the
Th?orie, and the framework within which it was generated.
During the years 1892-94, Duhem took up the fight against the
basic positivist empirical notion of science and against Anglo-Saxon
'modelism', which was still encountering noticeable diffidence among
the French. From those years to the year 1906, the year of publication
of the Th?orie, a number of riotous overlapping events considerably
changed the French scientific and cultural scene. A series of upsetting
experimental discoveries and an equally surprising sequence of theoreti
cal elaborations (especially tied to Lorentz's theories) imposed a realist
ically interpreted atomic theory, together with Maxwell's electromag
netism, upon the younger French scientists such as Perrin and Lange vin.
These events diffused the theme of scientific crisis.
The victory of atomism and electromagnetism meant victory - or at
least seemed to -
for that modelism Duhem thought already defeated.
These scientific events accompanied and even favored changes of great
importance in the French philosophical panorama, which was character
ized by the ever-increasing success of instrumentalistic, anti-intellectual,

This content downloaded from 195.220.213.10 on Tue, 16 Apr 2013 05:40:15 AM

All use subject to JSTOR Terms and Conditions
390 ROBERTO MAIOCCHI

and subjective concepts. Beginning in the 1890s, with the explosion of

the celebrated debate on the 'bankruptcy of science', French philosophy
was deeply marked by an impetuous blossoming of anti-intellectual
currents, such as Bergsonianism and modernism, radical conventional
ism, and varied forms of spiritualism. To use Fouill?e's famous words,
this period saw the "revolt of the heart against the intellect". Modelism
and instrumentalism, English physics, atomism and exasperated con
ventionalism, the crisis of science, anti-materialism and spiritualistic
skepticism seemed to form a thick web destined to surround and suffo
cate the model of scientific rationalism elaborated by Duhem in the
1890s.
To fight these foes, Duhem published a series of works in the early
1900s. The Th?orie represents the ultimate battle of this campaign. In
1902, he criticized electromagnetic theory very harshly in Les Th?ories
?lectriques de J. Maxwell. During the same year he attacked atomism
with Le mixte et la combinaison chimique. In 1903, with L'?volution de
la m?canique, he confronted the more generalized critique of the 'mod
elist' approach in its diverse historical variants. And finally, in 1904,
Duhem started publishing for the Revue de Philosophie a series of
articles which were eventually collected (with some additions) in 1906
to form La Th?orie physique. Here he fought the conventionalism
then in style on the epistemological level. The Th?orie physique was,
therefore, not at all a book opposing the positivism of the 1800s, in the
name of the new century's revolution in physics. It was a work against
the emerging novelties intended to demonstrate that the criticisms
brought a decade earlier against the positivist conception of science
need not give way to the skeptical conclusions that seemed to follow
directly from these novelties.
An analysis of the text of the Th?orie confirms the interpretation
which has led me to give the history of this work's genesis. For reasons
of space, it is impossible to carry out a detailed analysis here, but some
indications may be given: on all the key problems of epistemology
(what is a scientific fact? what is a law? how does one choose
theories?), Duhem clearly takes a position critical of the main conven
tionalists, primarily Poincar? and Le Roy, and fights against their
supposed solutions. I will briefly consider some examples.
Le Roy had given a rather strong subjectivist interpretation of the
'scientific fact', starting with the analyses made by Duhem in the 1890's
and maintaining that, due to the ineliminable theoretical components

This content downloaded from 195.220.213.10 on Tue, 16 Apr 2013 05:40:15 AM

All use subject to JSTOR Terms and Conditions
 A BOOK AGAINST CONVENTIONALISM 391

present in every experience, the 'scientific fact' is to a certain extent

'created' by the subject (Le Roy 1899, p. 516). Poincar? had retorted
by trying to subdue this radical subjectivism, maintaining that what the
scientist creates is the language with which we ask nature questions,
and it is then nature's task to give the answers (Poincar? 1905, p. 266
et
seq.).
Duhem argues at length even against Poincar?'s mitigated version
of conventionalism. The theme of science understood as a well-made

language is certainly not new: from Condillac and Lavoisier, through

the Ideologues, it had gone through positivism and had almost become
commonplace. It was directly connected to a depreciation of theories,
reducing them to the role of dictionaries which, through obviously
conventional rules, allowed the scientist to translate experience into
language. The
view had, in fact, been emphasized by radical conven
tionalists like D'Adh?mar (D'Adh?mar 1904). Against these general
positions Duhem emphasizes (Duhem 1906, p. 266) that science differs
from other languages as to its terms, just because they are defined
within a theoretical
context, stabilizing multiple interconnections in a
network of relationships between term and term, concept and concept,
not to mention relationships among some terms and groups of phenom
ena. A scientific fact is not differentiated from a nonscientific fact only
because it is expressed in a language resulting from customs known only
by a small group of people (Poincar?'s thesis). Its main characteristic is
that of belonging, by virtue of the theories that we use to express it,
to an intricate network of relationships with theoretical terms and with
a multitude of other scientific facts. When we translate a raw fact into
a scientific fact, we do not simply construct a proposition using the
expressions of a language provided with conventional rules known by
a small group of people (the scientists), we do much more. We insert
that fact in a sequential scheme, including other facts, and we recognize
relationships among phenomena. However, the linguistic translation of
the raw fact to the scientific fact is not simply made by choosing the
rules of translation freely and conventionally. It is guided by the theo
ries allowed at a given historical moment, and the result of the trans
lation work is, therefore, not invented by the scientist; it is the result
of history. It depends upon the level that science has reached at a given
historical moment. Science as a means of human expression is, in fact,
a language, but a language radically different from all others.

This content downloaded from 195.220.213.10 on Tue, 16 Apr 2013 05:40:15 AM

All use subject to JSTOR Terms and Conditions